LVS环境搭建及图形化管理(Pacemaker管理+heartbeat心跳+ipvs(DR)+ldirector健康检查及转发)

来源:互联网 发布:绘制平面图的软件 编辑:程序博客网 时间:2024/05/14 22:12
LVS环境搭建及图形化管理(Pacemaker管理+heartbeat心跳+ipvs(DR)ldirector健康检查及转发)
        很久没有正儿八经的写技术博客了,希望各位看官海涵。接下来,直接进入主题。
为了使用一个名叫LCMC(好像叫Linux Cluster Managerment Console)的lvs图形化管理界面,我这次lvs的主角确定为heartbeat(基本停止开发,只进行维护)。之前在公司的生产环境进行试验,所以LCMC可以直接连接公网进行wget和yum安装,所以一切都显得非常便捷,基本一个主备模式的lvs只需要几分钟就轻松搞定。为了防止试验环境对生产网的干扰,只得把主机放到测试环境中去。由于测试环境无法直接访问公网,所以花了一天把所需的安装包都收集好。好了,开始安装。
1.系统安装
操作系统:RHEL5.8 X86_64 版本
内核版本: 2.6.18-308.el5

2.配置

借图一用


服务器及IP为
虚拟IP(VIP) :                              eth0:192.168.29.100                  eth1:xxx.xxx.xxx.45/46
主LVS(ds1.lvs.stefentang):        eth0:192.168.29.45                  eth1:xxx.xxx.xxx.45
备LVS(ds2.lvs.stefentang):        eth0:192.168.29.46                  eth1:xxx.xxx.xxx.46
实机1(rs1.lvs.stefentang):       eth0:192.168.29.47(45)  eth0:无
实机2(rs2.lvs.stefentang):       eth0:192.168.29.48(46)  eth0:无

3.软件安装
为了便于日后集成到操作系统的安装包中去,都采用rpm包安装(所需安装包稍后上传)
在ds1和ds2上分别安装:
[root@ds1 ~]# rpm -ivh --force *.rpm
[root@ds2 ~]# rpm -ivh --force *.rpm
执行结果:
Preparing...                ########################################### [100%]   1:libtool-ltdl           ########################################### [  2%]   2:net-snmp-libs          ########################################### [  4%]   3:libesmtp               ########################################### [  5%]   4:openhpi-libs           ########################################### [  7%]   5:OpenIPMI-libs          ########################################### [  9%]   6:lm_sensors             ########################################### [ 11%]   7:libqb                  ########################################### [ 13%]   8:imake                  ########################################### [ 14%]   9:drbd83                 ########################################### [ 16%]  10:bzip2                  ########################################### [ 18%]  11:pacemaker-debuginfo    ########################################### [ 20%]  12:openais-debuginfo      ########################################### [ 21%]  13:heartbeat-debuginfo    ########################################### [ 23%]  14:glib2-devel            ########################################### [ 25%]  15:cluster-glue-debuginfo ########################################### [ 27%]  16:autoconf               ########################################### [ 29%]  17:zlib-devel             ########################################### [ 30%]  18:libxml2-devel          ########################################### [ 32%]  19:automake               ########################################### [ 34%]  20:libtool-ltdl-devel     ########################################### [ 36%]  21:perl-TimeDate          ########################################### [ 38%]  22:openib                 ########################################### [ 39%]  23:libsepol-devel         ########################################### [ 41%]  24:libselinux-devel       ########################################### [ 43%]  25:libgpg-error-devel     ########################################### [ 45%]  26:libgcrypt-devel        ########################################### [ 46%]  27:keyutils-libs-devel    ########################################### [ 48%]  28:e2fsprogs-devel        ########################################### [ 50%]  29:krb5-devel             ########################################### [ 52%]  30:openssl-devel          ########################################### [ 54%]  31:libesmtp-devel         ########################################### [ 55%]  32:libxslt-devel          ########################################### [ 57%]  33:libtool                ########################################### [ 59%]  34:bzip2-devel            ########################################### [ 61%]  35:kmod-drbd83            ########################################### [ 63%]  36:libqb-devel            ########################################### [ 64%]  37:resource-agents-debugin########################################### [ 66%]  38:cluster-glue-libs      ########################################### [ 68%]  39:cluster-glue           ########################################### [ 70%]  40:libibverbs             ########################################### [ 71%]  41:resource-agents        ########################################### [ 73%]  42:corosync               ########################################### [ 75%]  43:librdmacm              ########################################### [ 77%]  44:corosynclib            ########################################### [ 79%]  45:cluster-glue-libs-devel########################################### [ 80%]  46:libnes                 ########################################### [ 82%]  47:corosynclib-devel      ########################################### [ 84%]  48:heartbeat-libs         ########################################### [ 86%]  49:pacemaker-libs         ########################################### [ 88%]  50:openaislib             ########################################### [ 89%]  51:heartbeat              ########################################### [ 91%]  52:openais                ########################################### [ 93%]  53:pacemaker              ########################################### [ 95%]  54:heartbeat-devel        ########################################### [ 96%]  55:openaislib-devel       ########################################### [ 98%]  56:pacemaker-libs-devel   ########################################### [100%]
4.配置HB(/etc/ha.d/ha.cf)可以由LCMC来完成,也可以手动编辑。
#集群中的节点不会自动加入autojoin    none#heartbeat会记录debug日志,如果启用use_logd,则此选项会被忽略debugfile   /var/log/ha-debug#记录所有non-debug消息,如果启用use_logd,则此选项会被忽略logfile    /var/log/ha-log#告诉heartbeat记录那些sysloglogfacility   local0#指定两个心跳检测包的时间间隔keepalive 1#多久以后心跳检测决定集群中的node已经挂掉deadtime   30#心跳包检测的延时事件,如果延时,只是往日志中记录warning日志,并不切换服务warntime  10#在heartbeat启动后,在多长时间内宣布node是dead状态,因为有时候系统启动后,网络还需要一段时间才能启动initdead  120#如果udpport指令在bcast ucast指令的前面,则使用哪个端口进行广播,否则使用默认端口udpport   694#设置使用哪个网络接口发送UDP广播包,可以设置多个网络接口bcast  eth1 eth1#设置在哪个网络接口进行多播心跳检测mcast eth0 239.192.0.0 694 1 0#设置使用哪个网络接口进行UDP单播心跳检测,在ds2上为10.0.0.45#ucast  eth1 10.0.0.46#在主节点的服务恢复后,是否把从节点的服务切换回来auto_failback off#告诉集群中有哪些节点,node名称必须是uname -n显示出来的名称,可以在一个node中设置多个节点,也可以多次设置node,每一个在集群中的node都必须被列出来node ds1.lvs.stefentang ds2.lvs.stefentang#设置ping节点,ping节点用来检测网络连接#ping 10.0.0.254#开启Pacemaker cluster manager,因为历史原因,次选项默认是off,但是应该保持该选项值为respawn。在设置为respawn默认自动使用以下配置#pacemaker  respawncrm respawncompression bz2compression_threshold 20traditional_compression on#默认配置文件中下面还有很多选项,由于暂时用不到所以暂时忽略

5.配置/etc/ha.d/authkeys(LCMC完成,authkeys文件属性必须是600,否则日志会报错。)

## generated by drbd-guiauth 11 sha1 9AO5FmCl9RwBZZLIepNERwSbMMsmgeiF

6.配置ldirectord(/etc/ha.d/ldirectord.cf)可以手动编辑,也可以拷贝ldirectord的share目录下的文件

## Sample ldirectord configuration file to configure various virtual services.## Ldirectord will connect to each real server once per second and request# /index.html. If the data returned by the server does not contain the# string "Test Message" then the test fails and the real server will be# taken out of the available pool. The real server will be added back into# the pool once the test succeeds. If all real servers are removed from the# pool then localhost:80 is added to the pool as a fallback measure.# Global Directiveschecktimeout=3checkinterval=1#fallback=127.0.0.1:80autoreload=yeslogfile="/var/log/ldirectord.log"#logfile="local0"emailalert="tanghuifeng@besttone.com.cn"emailalertfreq=3600emailalertstatus=allquiescent=yes# Sample for an http virtual servicevirtual=192.168.29.100:80    real=192.168.29.47(45):80 gate    real=192.168.29.48(46):80 gate#    real=192.168.6.6:80 gate#     fallback=127.0.0.1:80 gate    service=http    request="health/health.html"    receive="running status: available"#    virtualhost=some.domain.com.au    scheduler=rr    #persistent=600    #netmask=255.255.255.255    protocol=tcp      checktype=negotiate      checkport=80     request="health/health.html"     receive="running status: available"     virtualhost=ds1/ds2.lvs.stefentang

测试心跳是否正常: tcpdump -i eth1 -n -p udp port 694

7.安装http服务器

lvs配置完后,配置真实服务器
[root@rs1 ~]# rpm -ivh /mnt/Server/apr-1.2.7-11.el5_6.5.x86_64.rpm[root@rs1 ~]# rpm -ivh /mnt/Server/postgresql-libs-8.1.23-1.el5_7.3.x86_64.rpm[root@rs1 ~]# rpm -ivh /mnt/Server/apr-util-1.2.7-11.el5_5.2.x86_64.rpm[root@rs1 ~]# rpm -ivh /mnt/Server/mailcap-2.1.23-1.fc6.noarch.rpm[root@rs1 ~]# rpm -ivh /mnt/Server/httpd-2.2.3-63.el5.x86_64.rpm[root@rs1 ~]# service httpd start

http服务80端口根目录添加一个health/health.html,内容为

[47(45)/48(46).DS1/2]running status: available

然后浏览器输入http://RIP/health/health.html ,能正确看到内容。
配置实机环路地址 lo:0(/etc/init.d/lvs_rs)
#!/bin/bash#chkconfig: 2345 80 90#description:lvs real server set lo:0 VIP addressVIP=192.168.29.100  #virtual ip    case "$1" in    start)    echo " start LVS of REALServer"    /sbin/ifconfig lo:0 $VIP broadcast $VIP netmask 255.255.255.255 up    echo "1" >/proc/sys/net/ipv4/conf/lo/arp_ignore    echo "2" >/proc/sys/net/ipv4/conf/lo/arp_announce    echo "1" >/proc/sys/net/ipv4/conf/all/arp_ignore    echo "2" >/proc/sys/net/ipv4/conf/all/arp_announce    ;;    stop)    /sbin/ifconfig lo:0 down    echo "close LVS Directorserver"    echo "0" >/proc/sys/net/ipv4/conf/lo/arp_ignore    echo "0" >/proc/sys/net/ipv4/conf/lo/arp_announce    echo "0" >/proc/sys/net/ipv4/conf/all/arp_ignore    echo "0" >/proc/sys/net/ipv4/conf/all/arp_announce    ;;    *)    echo "Usage: $0 {start|stop}"    exit 1    esac
然后
chmod u+x lvs_rs
/etc/inid.d/lvs_rs start

验证ldirectord是否做健康检查可以尝试查看httpd的access日志文件如(访问源从29.45变成29.46期间LVS做了一次切换):

192.168.29.45 - - [08/Aug/2013:07:32:16 +0800] "GET /health/health.html HTTP/1.1" 200 372 "-" "libwww-perl/5.805"192.168.29.45 - - [08/Aug/2013:07:32:17 +0800] "GET /health/health.html HTTP/1.1" 200 372 "-" "libwww-perl/5.805"192.168.29.46 - - [08/Aug/2013:07:33:12 +0800] "GET /health/health.html HTTP/1.1" 200 372 "-" "libwww-perl/5.805"192.168.29.46 - - [08/Aug/2013:07:33:13 +0800] "GET /health/health.html HTTP/1.1" 200 372 "-" "libwww-perl/5.805"

开始怎么也想不通,为啥切换动作差不多花了整整55秒,后来发现DNS配置有问题,干脆把DNS解析注释掉(修改/etc/resolv.conf文件):

192.168.29.45 - - [17/Aug/2013:02:46:37 +0800] "GET /health/health.html HTTP/1.1" 200 364 "-" "libwww-perl/5.805"192.168.29.45 - - [17/Aug/2013:02:46:38 +0800] "GET /health/health.html HTTP/1.1" 200 364 "-" "libwww-perl/5.805"192.168.29.46 - - [17/Aug/2013:02:46:55 +0800] "GET /health/health.html HTTP/1.1" 200 364 "-" "libwww-perl/5.805"192.168.29.46 - - [17/Aug/2013:02:46:59 +0800] "GET /health/health.html HTTP/1.1" 200 364 "-" "libwww-perl/5.805"

并且修改heartbeat的以下三个参数后,就缩短到17秒,估计也不能改得太小,造成误判。

warntime  5deadtime  10initdead 30


接下来都可以在LCMC端操作了,稍后配图。

8.查看配置结果

[root@ds1 ~]# crm_monAttempting connection to the cluster...============Last updated: Thu Aug  8 06:59:53 2013Stack: HeartbeatCurrent DC: ds1.lvs.stefentang (a4f7c50b-823d-4434-bb1d-850aa18918b2) - partition with quorumVersion: 1.0.12-unknown2 Nodes configured, unknown expected votes2 Resources configured.============Online: [ ds1.lvs.stefentang ds2.lvs.stefentang ]res_IPaddr2_VIP (ocf::heartbeat:IPaddr2):       Started ds1.lvs.stefentangres_ldirectord_1        (ocf::heartbeat:ldirectord):    Started ds1.lvs.stefentang

[root@ds2 ~]# crm_monAttempting connection to the cluster...============Last updated: Thu Aug  8 07:01:15 2013Stack: HeartbeatCurrent DC: ds1.lvs.stefentang (a4f7c50b-823d-4434-bb1d-850aa18918b2) - partition with quorumVersion: 1.0.12-unknown2 Nodes configured, unknown expected votes2 Resources configured.============Online: [ ds1.lvs.stefentang ds2.lvs.stefentang ]res_IPaddr2_VIP (ocf::heartbeat:IPaddr2):       Started ds1.lvs.stefentangres_ldirectord_1        (ocf::heartbeat:ldirectord):    Started ds1.lvs.stefentang

最后用ab并发1000,10000连接,压了一下,查看了分发情况如下:

IP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags  -> RemoteAddress:Port           Forward Weight ActiveConn InActConnTCP  192.168.29.100:http rr  -> 192.168.29.45:http           Local   1      3          5071  -> 192.168.29.46:http           Route   1      4          4930
Server Software:        Apache/2.2.3Server Hostname:        192.168.29.100Server Port:            80Document Path:          /is/checkNumber.doDocument Length:        1336 bytesConcurrency Level:      1000Time taken for tests:   68.166 secondsComplete requests:      10000Failed requests:        9883   (Connect: 0, Receive: 0, Length: 9883, Exceptions: 0)Write errors:           0Total transferred:      16318053 bytesHTML transferred:       13448053 bytesRequests per second:    146.70 [#/sec] (mean)Time per request:       6816.590 [ms] (mean)Time per request:       6.817 [ms] (mean, across all concurrent requests)Transfer rate:          233.78 [Kbytes/sec] receivedConnection Times (ms)              min  mean[+/-sd] median   maxConnect:        1    7  42.7      5    3014Processing:    81 6228 1490.0   6244   11462Waiting:       33 3396 2218.5   3159   10526Total:         86 6235 1490.7   6250   11467

心血来潮,测了一下ab的极限,呵呵:

==================================10000=====================================InActConn连接数上不去了,保持在2600左右IP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags  -> RemoteAddress:Port           Forward Weight ActiveConn InActConnTCP  192.168.29.100:http rr  -> 192.168.29.45:http           Local   1      4996       2563  -> 192.168.29.46:http           Route   1      5003       2557==================================20000=====================================ActiveConn连接数上不去了,保持在7500左右,ab压力机报错:apr_socket_connect(): 由于系统缓冲区空间不足或队列已满,不能执行套接字上的操作。(730055)IP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags  -> RemoteAddress:Port           Forward Weight ActiveConn InActConnTCP  192.168.29.100:http rr  -> 192.168.29.45:http           Local   1      7624       2188  -> 192.168.29.46:http           Route   1      7571       2186